Research On Stop Behavior Of Engine Based On Data Mining Theory

In recent years, as the development of economy of China and improvement living standard of our people,more and more families would like to buy a private car as their transportation, which made the numberof vehicles in our country keep rising, and automobile industry of China got the valuable chance to growfast. All kinds of vehicles brought many sorts of conveniences to our production and living, but, at thesame time, domestic energy security and the environment pollution problems, especially pollutionconcentration problem in urban area also came out increasingly. In these social problems environment, anew engine starting technique which can improve fuel consumption and emissions of engines graduallycame into vision field of scientific research groups who are responsible for developing vehicle dynamicplatform. It is the Instantaneous Reverse Direct Start&Stop System based on gasoline direct injection(GDI) engines.However, this kind of starting technology presented a high demand on stop-position of enginepiston, which made the implementation and promotion of it be difficult to realize. At present, domesticand international research institutions don’t have clear and detailed report on the stop process of engine,due to the limitations from traditional measurement equipments and many aspects. In order to breakthrough those technical difficulties, this thesis designed the measurement system of engine pistonposition in all-operating condition and its measuring method, and use this measurement system andother related test equipments to research on stop behavior of engines in every test condition by usingdirect injection gasoline engine as the research object. This thesis presented the definition and researchmethod of stop behavior of engine, and summarized the characteristics of stop behavior of engine.“Pumping effect” which is an important effect in the process of stop behavior was found. Non-linearregression model and radial basis function neural network model were established by the data from stopengine experiments. Non-linear regression model can predict the expectation of piston position under the single-conditions, and radial basis function neural network model can predict piston postion under thecompound-conditions. Therefore, This thesis presented the general law about the influence of coolingwater temperature, initial stop speed and stop load on piston position of engine.The main work includes:Design and development of measurement system of engine piston position in all-operatingconditions. Because the disadvantages of traditional sensors can not measure parameters of engine atlow speed, this thesis adopted photoelectric sensor instead of current magnetoelectric sensor, whichmade piston position of engine, phase and speed and so on be measured accurately from highest speeduntil crankshaft static. Therefore, this measurement system provided a research platform for stopbehavior of engine. Define stop behavior of engine and propose the method to research and describestop behavior of engine. Due to using measurement system of engine piston position in all-operatingconditions, the process of engine stopping integrated into engine operating conditions. Therefore, thisthesis defined the process of engine stopping as stop behavior of engine. Due to no current researchachievement about stop behavior of engine, this thesis first proposed the method to describe andresearch, and divided mass of test data into two types. One type is result type data, another is processtype data. So the stop behavior of engine can be analyzed from different aspects.Establish engine test bench for stop behavior of engine. This thesis designed the test bench forengine and confirmed those test conditions for stop behavior of engine. According to the arrangement ofengine test bench, the measurement system of engine piston position in all-operating conditions wasinstalled, and initialization and calibration can be completed. Besides, KISTLER cylinder pressuresensor and AVL combustion analyzer were also installed on this engine test bench in order to measurepressure in the cylinder. Based on this engine test bench, the real-time measurement for piston position,cylinder pressure, speed, temperature of cooling water in engine and other parameters can be realized.Therefore, the whole stop behavior can be known clearly.Based on a mass of procedural data from engine stop experiment, the special reversal phenomenonof crankshaft will appear from stopping injecting fuel and sparking to being totally static state ofcrankshaft, even normal rotation again after reversal phenomenon will appear. Therefore, the crankshaftof engine will experience “normal rotation-reverse” or “normal rotation-reverse-normal rotation again”,which is the important feature of stop behavior of engine. Through the analysis of cylinder pressure data during the engine stop process, cylinder pressure didnot descent but increase when there are no working medium work was found. This kind of phenomenonwas named “pumping effect” in this thesis. After analyzing the exhaust process of engine, this thesispresented the main reason of “pumping effect” was found, which is that supercritical exhaust disappearduring the exhaust process. Based on the gas state equation, exhaust effect will be weaken and intakeeffect will still be the same, so cylinder pressure will be increased cycle by cycle during the engine stopprocess. Therefore, to some extent, the “pumping effect” made the function of building cylinder pressureby traditional starter was transferred to engine stop process. It also further proved that the instantaneousreverse direct start&stop system based on direct injection gasoline engine is feasible.Nonlinear regression model was established with SPSS software. This thesis will use the CurveSimulation Module of SPSS software to do regression analysis between the expectations of pistonpostion and stop engine conditons, and then according to the decision factorR2to choose appropriatemodel for engine stop behavior make predictions, and stop behavior of engine can be analyzed undersingle-condition, that meant the expectation of piston postions were predicted if only one codition waschanged. Radial basis function (RBF) neural network was established with Matlab software, and RBFneural network was trained by all data from engine stop experiment. The change of piston postion waspredicted under compound-condition, therefore, the research on stop position of piston can begeneralized from line conditions from surface conditions.Using radial basis neural network model to analyze engine stop behavior can know the general rule:(1) If initial stop speed decrease and stop load increase at the same time, the angle between the pistonpostion and the top dead center will be decreased. Otherwise, if initial stop speed increase and stop loaddecrease at the same time, the angle between the piston postion and the top dead center will be increased;(2) When stop load did not change, the angle between the piston postion and the top dead will beincreased along with the increase of initial stop speed;(3) When initial stop speed did not change, theangle between the piston postion and the top dead will be decreased along with the increase of stop load;(4) If stop load and initial stop speed tend to be in the medium area of measurement area, angle betweenthe piston postion and the top dead will fluctuate in the local area.(5) The influence of cooling watertemperature on engine stop behavior is lesser. Rregardless of cooling water temperature changing, ingeneral, engine stop behavior will comply with the above rules. Those conditions which can help engine piston stop in the target area will be found out based on the above law.